Research On Intelligent Traffic Flow Forcasting And Route Guidance Algorithm Based On Hadoop Platform

With developing rapidly of the complex traffic systems,although related departments have invested a lot of money and resources in all sorts of traffic facilities construction,they still can't satisfy people's travel demand.For such complex transportation system,in order to improve the searching efficiency of intelligent transportation systems,reduce the search scope,feedback network information in a shorter period of time,shorten the "unnecessary waiting time" of outing for users,It is necessary to study the prediction of short-term traffic flow and the algorithm of route induction.However,to solve the above problem,the most important thing is to improve the efficiency of short-term traffic flow forecast and path-induced algorithm.From the perspective of short-term traffic flow prediction algorithm and path-induced algorithm,the effectiveness of traffic management is directly related to the prediction,the induced precision and the efficiency of the algorithm.But,accuracy and efficiency are negative correlated,the higher of the precision of the algorithm,the more complex of its logic or the heavier of the calculation.so,computing takes longer,lower efficiency,and completely lose practicability when it is serious.In terms of short-term traffic flow prediction,the thesis analyses the common short-term traffic flow prediction algorithm,and points out the advantages and disadvantages of various algorithms as well as the using range,due to the mathematical model of the BP neural network algorithm has characteristic of rigor and at the same time,with good self-learning ability,good fault tolerance and good generalization,so the BP neural network algorithm is used to study the short-term traffic flow.But the BP neural network algorithm use the static gradient descent methods to optimize the network weight and threshold value,so make the BP neural network algorithm have some limitations,such as stable,slow convergence rate,easy to achieve local minimum.In order to overcome the above defects,the thesis uses the modified genetic algorithm to optimize the BP neural network prediction model in the forecast of short-term traffic flow.As a global scope of the search algorithm for genetic algorithm,through simulation in the process of the genetic characteristics of reproduction,crossover and mutation genetic factor,the individual best unceasingly,will eventually get the optimal solution as the initial value of neural network algorithm.But the complex diversity of traffic flow data makes genetic algorithms in the process of searching optimal solution possible loss,leading to premature convergence,it reduces the accuracy of short-term traffic flow prediction.In order to overcome the above defects,the chaotic genetic algorithm(CGA),which has introduced in the genetic algorithm,and extremely matched to the transient traffic flow motion.Its core thought is mainly introducing chaos state in optimization variables,and the chaotic motion of traverse range "extension" to stay in the scope of optimized variables,global organic search,so it can avoid premature into local optimal solution,and finally the optimal solution is obtained by constantly improved.Then the initial weight and threshold value of BP neural network are initialized with the optimal solution,which increases the efficiency and accuracy of the forecast of short-term traffic flow.The thesis also verified the improved performance of the improved algorithm.In route guidance,the thesis also has studied the route guidance algorithm which is commonly used,and has analyzed the advantages and disadvantages of each algorithm,as well as the using range,the ant colony algorithm with intelligent search,can achieve the goal of global optimization,the robustness,self-organizing,parallelism performance is very outstanding,and is suitable for complex nonlinear system,the traffic so the ant colony algorithm is adopted to study route guidance.Of course,any algorithm has its own limitations and shortcomings,the paper aiming at the flaws of the ant colony algorithm,respectively of the ant colony algorithm state transition rule and pheromone update rules was improved.So as to reduce the search of an invalid path for the user's travel,and is able to select the optimal path from a combination of factors.In the study of this thesis,to meet the requirement of the practical requirement of shortterm traffic flow prediction and path induction algorithm,make full use of the advantages of cloud computing platform in data preservation and parallel processing,combine with the Hadoop platform,design and realization of Map Reduce for the improved BP neural network algorithm and ant colony algorithm were carried out,Successfully designed new short-term traffic forecast and path induction method,find the right balance between the accuracy and efficiency of In prediction and induction,greatly strengthen the two algorithm in practical performance,and the performance and practicability of the algorithm have been verified in the experiment.